Apps, Agents, and Improvisation: Ensemble Interaction with Touch-Screen Digital Musical Instruments

Abstract

This thesis concerns the making and performing of music with new digital musical instruments (DMIs) designed for ensemble performance. While computer music has advanced to the point where a huge variety of digital instruments are common in educational, recreational, and professional music-making, these instruments rarely seek to enhance the ensemble context in which they are used. Interaction models that map individual gestures to sound have been previously studied, but the interactions of ensembles within these models are not well understood. In this research, new ensemble-focussed instruments have been designed and deployed in an ongoing artistic practice. These instruments have also been evaluated to find out whether, and if so how, they affect the ensembles and music that is made with them.

Throughout this thesis, six ensemble-focussed DMIs are introduced for mobile touch-screen computers. A series of improvised rehearsals and performances leads to the identification of a vocabulary of continuous performative touch-gestures and a system for tracking these collaborative performances in real time using tools from machine learning. The tracking system is posed as an intelligent agent that can continually analyse the gestural states of performers, and trigger a response in the performers' user interfaces at appropriate moments. The hypothesis is that the agent interaction and UI response can enhance improvised performances, allowing performers to better explore creative interactions with each other, produce better music, and have a more enjoyable experience.

Two formal studies are described where participants rate their perceptions of improvised performances with a variety of designs for agent-app interaction. The first, with three expert performers, informed refinements for a set of apps. The most successful interface was redesigned and investigated further in a second study with 16 non-expert participants. In the final interface, each performer freely improvised with a limited number of notes; at moments of peak gestural change, the agent presented users with the opportunity to try different notes. This interface is shown to produce performances that are longer, as well as demonstrate improved perceptions of musical structure, group interaction, enjoyment and overall quality.

Overall, this research examined ensemble DMI performance in unprecedented scope and detail, with more than 150 interaction sessions recorded. Informed by the results of lab and field studies using quantitative and qualitative methods, four generations of ensemble-focussed interface have been developed and refined. The results of the most recent studies assure us that the intelligent agent interaction does enhance improvised performances. Show more